Common ancestry or environmental trait filters: cross-continental comparisons of trait-habitat relationships in tropical anuran amphibian assemblages

Aim To investigate whether traithabitat relations in biological communities converge across three global regions. The goal is to assess the role of habitat templets in shaping trait assemblages when different assembly mechanisms are operating and to test whether traithabitat relations reflect a common evolutionary history or environmental trait filters. Location Guiana Shield, South America; Upper Guinea Forest Block, West Africa; Borneo rain forests, Southeast Asia. Methods We compared large anuran amphibian data sets at both the regional and cross-continental scale. We applied a combination of three-table ordinations (RLQ) and permutation model-based multivariate fourth-corner statistics to test for traithabitat relationships at both scales and used phylogenetic comparative methods to quantify phylogenetic signal in traits that enter these analyses. Results Despite the existence of significant traithabitat links and congruent trait patterns, we did not find evidence for the existence of a universal traithabitat relationship at the assemblage level and no clear sign for cross-continental convergence of traithabitat relations. Patterns rather varied between continents. Despite the fact that a number of traits were conserved across phylogenies, the phylogenetic signal varied between regions. Traithabitat relations therefore not only reflect a common evolutionary history, but also more recently operating environmental trait filters that ultimately determine the trait composition in regional assemblages. Main conclusions Integrating traithabitat links into analyses of biological assemblages can enhance the predictive power and general application of species assembly rules in community and macroecology, particularly when phylogenetic comparative methods are simultaneously applied. However, in order to predict trait composition based on habitat templets, traithabitat links cannot be assumed to be universal but rather have to be individually established in different regions prior to model building. Only then can direct trait-based approaches be useful tools for predicting fundamental community patterns.